The merging of those 3 domains is very interesting. It is beginning to be referred to with its own name, Cognitive Science.
Turing's paper will clear up some of the problems, but he is guilty as charged. He argues that we can not find a rigorous definition of intelligence, and he replaces it with the Turing test (he calls it the imitation game). However, he does explain in his paper why this replacement is valid.
I do understand your point, and several people have made it in the past. However, Artificial Intelligence is an empirical domain. We may never understand everything about how the brain works, yet that doesn't mean we can never approach it. We still do not understand the "principles of flight" (people still disagree about how an airplane wing works). To follow your argument would imply that we don't have machines that can fly; they simply "mimic" flight.
I think that this example might not be a real breakthrough, but the approach is. It is quite likely that intelligence is an emergent behavior. Trying to understand it by following the Boolean dream (Rules for Thought), as you suggest, is a chimera.
I am a logic and computation major, considering cognitive science as a second major. So yes, I do know all of those things. I suggest rather than swallow Penrose's line (I have read some of his work) that you read some of the many people who disagree with him. His views are not very popular for good reason. A good place to start is with Daniel Dennet or Douglas Hofstader. They go over the subject in a highly accesible way.
Have you ever read Turing's paper? He addresses most of the objections people bring up again and again. The problem with your analogy is that con-artists are intelligent. What Turing proposes is that if you can be a con-artist, you must be intelligent.
Language is the hardest problem that we know of. Human language is productive, which means that the number of well-formed sentences is not finite. In set theory terms, this means that the members of the set of "sentences in the English language" can not be created algorithmicly.
What this all means is that no chatterbot-type program could ever fool a human for long. Humans recognize and create entirely new sentences all of the time. A computer program that could do this consistently at a human level would need to have a complex structure. In particular, it would need to have some semantic understanding, not just manipulate sentences according to simple rules. This, to me, would be entirely reasonable to call intelligent.
Absolutely. It is ridiculous to imagine that an intelligent system could ever emerge from an arrangment of simple chemicals or the firing of neurons. These people that claim that a "human", no matter how convincing its responses, could be intelligent are crazy. By all means, lets make interesting biological constructs, but these assertions of "strong intelligence" are complete malarkey.
"We can have a personal assistant, a slave, a friend who doesn't really suffer by being delegated these tasks," [Dr. Anat Treister-Goren] said. "
It is doubtful that intelligence could exist in anything content to be saddled with such tasks. A so-called "expert" system that doesn't pretend to be intelligent would be better suited.
True, it would be difficult to emulate that many neurons. There have been some attempts to create neural networks in hardware, and there are also several specialized boards on the market that connect to a traditional serial computer. This does sound like it's in software though.
It wouldn't be a genetic algorithm. Genetic algorithms iterate over several generations of programs. This sounds like a single program that learns. It seems more likely that it uses a combination of a Hebbian learning algorithm (for neural networks) and some symbolic approach.
C was first designed by Dennis Ritchie for use with UNIX on DEC PDP-11 computers. The language evolved from Martin Richard's BCPL, and one of its earlier forms was the B language, which was written by Ken Thompson for the DEC PDP-7.
If we solved language processing and curiosity, we would probably have an intelligence. Language processing alone is believed to be the most difficult problem in cognitive science.
I agree, the problems that you point out are real. Recurrent neural networks address some of the problems with NNs, and a real network may need to be implemented in hardware to gain the benefits of analog operation.
I suspect that we will have to program in some basics as well, as you say. The benefits of neural networks have a lot to do with their heuristic ability: our thinking is full of analogies and fuzzy reasoning. The cost is that it may turn out to be impossible to learn abilities such as language (as Chomsky argues). Genetic algorithms may stumble across the right set of initial weights for language processings, but symbolic manipulation will probably help out there.
Broadcast 2000 is one package available - i've played a bit with an early version, and the interface wasn't great, but it had the basics of video and sound editing. Comparable to consumer level software bundled with your camera, not yet close to a Premiere (or even a Vegas Video). A nice feature is that it is a multi-track editor.
I don't know of any compositing software for Linux that is Free, and the available modeling/animation software is all very limited. However, there's a great renderer in POVRAY.
I don't mean to be a downer, but there were several factual mistakes in this article. DeAnne Dewitt's article skims over the information in the popular media. For instance, the school for robotics that she looks at was MIT, while the most research in this area is taking place at Carnegie Mellon. She has several arbitrary categories, while a natural division in current AI would probably be "symbolic" vs "connectionist".
The classical approach to AI was the symbolic, which grows out of Turing's work. Allen Newell and Herbert Simon of Carnegie Mellon University (not even mentioned) were the foremost promoters of this approach, which they called physical symbol systems. Other early AI pioneers include Marvin Minsky, who should probably be mentioned in any article on AI (but was not in this one).
The author barely mentions the neural network, or connectionist approach. These did not start with the PDP group, as she suggests, but with Frank Rosenblatt of Cornell, with Perceptrons. The most exciting research in this area deals with recurrent neural networks, which exhibit chaotic behavior. This is where I personally think that real intelligence could come from, because it is a more natural model of our brain's operations. The foremost researcher in this domain is Hava Siegelmann of Technion Institute in Haifa, Israel. She promotes the idea of analogical systems, which she has proven have more theoretical power than the Turing machine model.
If you want an introduction to AI, skip this article. A good place to start might be a scientific journal, or the comp.ai faq. Her resources are not very good either, so don't bother.
I don't know the details, but shouldn't it be possible to add the DeMuDi source lines to your sources file in Debian? In principle, since this is a Debian based distribution, all one would need to do is a simple apt-get. The great thing about making this a separate distribution is that if all you want is music at first, you start with DeMuDi. Want to upgrade to the whole Debian distribution? Change your sources.list to point to a Debian mirror. I don't know if this is how the DeMuDi maintainers are thinking, but it should be.
I think this is cool stuff, and means that the main Debian distributors don't need to put as much effort into maintaining these specialized packages. It is in keeping with the specialized Debian Jr. distribution that is currently in progress, as well.
Keep up the good work, and I for one am completely in favor of it.
All Netscape versions have plenty of annoying bugs and inconsistencies, in my experience more than IE. It may be that Netscape 6 is more compliant, but I have had more problems with Netscape 6's implementation of CSS than with Netscape 4.7's. I guess it isn't the number of features that are supported, but how consistently and which particular ones are commonly used. I like to stick only to 100% CSS compliant code, and I hate the workarounds that I have to use.
FYI, I believe that the first completely CSS compliant browser was IE/Mac.
Deceptive business practices will always be illegal. Capitalism breaks down when we don't have accurate information, and even most Republicans agree that preserving the integrity of information flow is a valid role for the government. Would you say that it was ok to sell a car that would kill the owner? Maybe you would say that if the car was bad, nobody would buy it. How do we find out it is bad? By the time we buy it, it's too late. That's where the FTC comes in.
That actually already exists, and is a feature of the file system, not the window manager. It's called a link. In Ext2FS, the link points to an inode, not a directory entry, so always remains current.
Slashdot Mirror
Shockwave is not the same as Flash. It is much harder to port because of the number of OS-specific calls it makes.
Turing's paper will clear up some of the problems, but he is guilty as charged. He argues that we can not find a rigorous definition of intelligence, and he replaces it with the Turing test (he calls it the imitation game). However, he does explain in his paper why this replacement is valid.
I think that this example might not be a real breakthrough, but the approach is. It is quite likely that intelligence is an emergent behavior. Trying to understand it by following the Boolean dream (Rules for Thought), as you suggest, is a chimera.
I am a logic and computation major, considering cognitive science as a second major. So yes, I do know all of those things. I suggest rather than swallow Penrose's line (I have read some of his work) that you read some of the many people who disagree with him. His views are not very popular for good reason. A good place to start is with Daniel Dennet or Douglas Hofstader. They go over the subject in a highly accesible way.
Language is the hardest problem that we know of. Human language is productive, which means that the number of well-formed sentences is not finite. In set theory terms, this means that the members of the set of "sentences in the English language" can not be created algorithmicly.
What this all means is that no chatterbot-type program could ever fool a human for long. Humans recognize and create entirely new sentences all of the time. A computer program that could do this consistently at a human level would need to have a complex structure. In particular, it would need to have some semantic understanding, not just manipulate sentences according to simple rules. This, to me, would be entirely reasonable to call intelligent.
Absolutely. It is ridiculous to imagine that an intelligent system could ever emerge from an arrangment of simple chemicals or the firing of neurons. These people that claim that a "human", no matter how convincing its responses, could be intelligent are crazy. By all means, lets make interesting biological constructs, but these assertions of "strong intelligence" are complete malarkey.
"We can have a personal assistant, a slave, a friend who doesn't really suffer by being delegated these tasks," [Dr. Anat Treister-Goren] said. "
It is doubtful that intelligence could exist in anything content to be saddled with such tasks. A so-called "expert" system that doesn't pretend to be intelligent would be better suited.
True, it would be difficult to emulate that many neurons. There have been some attempts to create neural networks in hardware, and there are also several specialized boards on the market that connect to a traditional serial computer. This does sound like it's in software though.
It wouldn't be a genetic algorithm. Genetic algorithms iterate over several generations of programs. This sounds like a single program that learns. It seems more likely that it uses a combination of a Hebbian learning algorithm (for neural networks) and some symbolic approach.
He designed a machine called the Collossus to break the German Enigma code (used for the U-Boat transmissions) for British Intelligence.
I'm one of those off campus people who will be changing DSL providers. Any tips now? You can send an email to me too - quinten@
Credit: Unix Unleashed
Actually, that's exactly where the name for C came from. I remember reading programming books that were surprised C++ wasn't named 'P'.
Get an OC3 if you can afford it.
And therefore the correct name should be 'P' or possibly 'L'. Why would it be D or E?
If we solved language processing and curiosity, we would probably have an intelligence. Language processing alone is believed to be the most difficult problem in cognitive science.
I suspect that we will have to program in some basics as well, as you say. The benefits of neural networks have a lot to do with their heuristic ability: our thinking is full of analogies and fuzzy reasoning. The cost is that it may turn out to be impossible to learn abilities such as language (as Chomsky argues). Genetic algorithms may stumble across the right set of initial weights for language processings, but symbolic manipulation will probably help out there.
I don't know of any compositing software for Linux that is Free, and the available modeling/animation software is all very limited. However, there's a great renderer in POVRAY.
The classical approach to AI was the symbolic, which grows out of Turing's work. Allen Newell and Herbert Simon of Carnegie Mellon University (not even mentioned) were the foremost promoters of this approach, which they called physical symbol systems. Other early AI pioneers include Marvin Minsky, who should probably be mentioned in any article on AI (but was not in this one).
The author barely mentions the neural network, or connectionist approach. These did not start with the PDP group, as she suggests, but with Frank Rosenblatt of Cornell, with Perceptrons. The most exciting research in this area deals with recurrent neural networks, which exhibit chaotic behavior. This is where I personally think that real intelligence could come from, because it is a more natural model of our brain's operations. The foremost researcher in this domain is Hava Siegelmann of Technion Institute in Haifa, Israel. She promotes the idea of analogical systems, which she has proven have more theoretical power than the Turing machine model.
If you want an introduction to AI, skip this article. A good place to start might be a scientific journal, or the comp.ai faq. Her resources are not very good either, so don't bother.
This is true - Chuck Schumer is far from being any New Yorker's favorite person. He is a true politician, and is looking for some limelight.
Check again; I saw it on the packages list.
I think this is cool stuff, and means that the main Debian distributors don't need to put as much effort into maintaining these specialized packages. It is in keeping with the specialized Debian Jr. distribution that is currently in progress, as well.
Keep up the good work, and I for one am completely in favor of it.
FYI, I believe that the first completely CSS compliant browser was IE/Mac.
The problem is when they lie about what they are doing. This has been well established through court precedent.
Deceptive business practices will always be illegal. Capitalism breaks down when we don't have accurate information, and even most Republicans agree that preserving the integrity of information flow is a valid role for the government. Would you say that it was ok to sell a car that would kill the owner? Maybe you would say that if the car was bad, nobody would buy it. How do we find out it is bad? By the time we buy it, it's too late. That's where the FTC comes in.
That actually already exists, and is a feature of the file system, not the window manager. It's called a link. In Ext2FS, the link points to an inode, not a directory entry, so always remains current.